Figure 1 shows earthquake activity along the three plate boundaries studied as well as within the Gorda plate. The three meet at what is called the Mendocino triple junction. The region is dominated by earthquakes with strike-slip mechanisms (yellow symbols). They occurred along the Mendocino transform fault near and to the northwest of the 1992 mainshock as well as along the Blanco transform fault. Shocks along the spreading Gorda ridge involve a predominance of normal faulting. Other earthquakes occurred within the downgoing Gorda plate where it is being subducted beneath the North American plate.

The southern Gorda plate is unusual in that many earthquakes occur within it with strike-slip mechanisms. It has been known since the early days of plate tectonics that the Gorda plate has been deformed during the last 10 million years [6] [7]. The deformed seafloor magnetic anomalies on the southeastern side of the Gorda plate result from compression as the Pacific plate converges upon it along the Mendocino transform (Figure 1). That transform has been moving northward for about the last 30 million years. The Gorda plate shrinks in size as the San Andreas fault system continues to grow to the northwest [6] [7]. Some North American-Pacific plate motion occurs to the east of the San Andreas fault (southeast of Figure 1) especially along the Maacama and Bartlett Spring faults.

Much of the northernmost San Andreas fault, described later, has been remarkably quiet even for microearthquakes.

Only the southernmost 100-km extent of the Cascadia subduction zone off the northwestern United States has been seismically active for large shocks during the last 100 years (Figure 1). It and the entire plate boundary farther north broke in a giant subduction zone shock in 1700 and in earlier events that were detected by paleoseismic studies [8]. Work on the 1992 Cape Mendocino earthquake may have the potential to understand better great shocks along that subduction zone to the north as far as British Columbia and possibly the norther end of the San Andreas fault. Separating earthquakes by their mechanism is a major step in identifying precursors to the 1992 mainshock.

In 1993 [9] published a description of several data sets for the 1992 mainshock. Rupture initiated onshore at a depth of 10.5 km and propagated up-dip and seaward. It damaged local towns, generated coastal uplift and produced a tsunami. Figure 2 shows the rupture zone inferred by [9] from changes in vertical deformation as a dashed rectangle.

The recomputed centroid moment-tensor, CMT, [1] places it near the northern end of the rupture zone as inferred by [9] and 30 km to its west (Figure 2). The original CMT produced in 1992 was located about 20 km to the south and 10 km to the east of the recomputed one [10] [11]. The revised dip of the CMT was increased from 9˚ to 16˚, resulting in Mw being revised downward from 7.15 to 6.91 [1]. Each CMT indicates a predominance of thrust faulting of shallow dip. That distinguishes the 1992 shock from the many of the strike-slip solutions in the region.

While the absolute location of the revised CMT is uncertain, its location suggests that rupture extended farther west than that deduced from geodetic data by

[9]. [12] used long-period seismic waves to deduce that the 1992 mainshock ruptured WSW from the initial short-period hypocenter. They place the rupture zone as far west as that determined from geodetic data as well as 5 km to its south but only in its southern half. I later discuss the volume of the zone that was largely quiet before the 1991 and 1992 earthquakes as another measure of the approximate size of the 1992 event.

Figure 2 illustrates patterns of forerunning and aftershock activity in an enlarged view near the 1992 mainshock. Forerunning activity from 1984 to the 1991 Honeydew earthquake in blue is largely concentrated either to the south or north of most of the 1992 aftershocks (green). Most activity between the 1991 Honeydew shock and the 1992 mainshock (red) occurs either along the Mendocino transform or within the Gorda plate but only up to 11 km west of the 1991 short-period hypocenter. Some of those events are aftershocks of the 1991 mainshock. Others are far enough west, however, that they can be interpreted as earthquakes triggered by the final stress buildup to the 1992 shock. It Is clear that the 1991 shock occurred very close to the 1992 rupture zone. Various events are portrayed separately later by their depth of focus.

Figure 3 illustrates the same earthquakes as in Figure 2 but as a function of

depth and longitude. The aftershocks fall largely into two groups—those shallower than 11 and ones deeper than 15 km. The aftershocks in red for the 1991 Honeydew earthquake define a plane dipping shallowly, which is consistent with its revised CMT solution (Table 1). Its aftershocks indicate that it ruptured westerly and up dip. Likewise, events after the 1992 mainshock indicate a similar rupture pattern but one that extended farther west and north as discussed later. The 12-km depth of the 1992 main event is the shallowest permitted by CMT processing. Its short-period depth was 10.5 km. The mainshock likely ruptured somewhere between depths of 6 and 11 km.

Figure 4 expands a similar view for depths only between 0 and 15 km. Aftershocks of the 1992 mainshock extended farther west than those for the 1991 event. The 1991 ones are also more concentrated in depth. Little activity occurred between depths of 0 and 4 km.

It is difficult to tell how far westerly rupture took place in the 1992 mainshock since the two largest aftershocks of Mw 6.5 and 6.6 occurred within one day.

Their depths (Figure 3), however, place them within a deeper zone of activity in the downgoing Gorda plate. Their mechanisms and rupture modelling [12] indicate one involved strike-slip motion on a northeasterly-striking zone and the other strike slip on a northwesterly-trending zone. While they were triggered by the1992 mainshock, their mechanisms and depths were decidedly different.

Figure 5 shows activity for depths of 0 to 14 km as a function of latitude and longitude. Aftershocks shallower than 14 km occurred between the Mendocino transform to the south and the CMT solution for the 1992 mainshock 17 km to the north. Aftershocks of 1992 were largely delimited to the north by the Bear River fault (BRF of Figure 5).

Most of the activity shown between those latitudes in Figure 2 was deeper than 14 km. It is clear that few forerunning shocks to 1992 of those depths took place north of the 1991 aftershocks and south of the 1992 CMT solution. That region can be described as a large asperity that broke in the 1992 mainshock. It denotes a region that probably slipped previously as a large thrust mainshock at an unknown date.

Very large mainshocks and their forerunning events are described in terms of the rupture of asperities of various sizes, i.e., relatively strong, well-coupled portions of plate interfaces. The term asperity is used extensively in the seismological and rock mechanics literatures. Some parts of plate boundaries consist of great asperities that are well coupled, i.e., largely locked, during the slow process of stress buildup to large earthquakes. Other parts, so called low-coupling zones (LCZ), often are identified as the sites of either moderate-magnitude forerunning

activity, smaller asperities or fault creep. Much forerunning activity as well as several slow-slip events described in the literature occurred in LCZ’s on the peripheries of great asperities world-wide [14] [15]. The pattern of forerunning activity to the 1992 mainshock was similar even though it likely occurred above the plate boundary.

Many of the larger forerunning shocks in Figure 2 occurred at depths below 14 km, i.e., deeper than that of the rupture zone of the main 1992 earthquake. Many other smaller forerunning events shown in Figure 6 took place at depths deeper than 14 km. Thus, they and all but one of the larger forerunning earthquakes occurred below the large asperity that broke in 1992. The quiet zone and its inferred large asperity were three-dimensional in shape.

Figure 7 shows known mechanisms of earthquakes with depths between 0 and 14 km from 1984 until the 1991 Honeydew shock plus three thrust mechanisms (two close together) from it in 1991 until the 1992 mainshock. An ellipse in red is drawn around the region of few forerunning thrust and reverse faulting events to approximately define the size of the coming 1992 mainshock. It is

about 20 km in diameter. Figure 5 and Figure 7 indicate that care must be used in defining quiet forerunning zones to large events so as to include depth as well as latitude, longitude and magnitude.

The 1992 mainshock has been attributed to slip either along the Gorda-North American plate boundary or along a similar-dipping surface a several kilometers above it [9]. [18] interpreted that it occurred along a blind thrust fault above the plate boundary. [19] reached a similar conclusion based on mapping the velocity ratio v_p/v_s.

Aftershocks in Figure 4 indicate the main rupture in 1992 took place on a shallow-dipping interface at depths between 7 to 9 km with perhaps slip on a sub-parallel fault at depths of 10 to 11 km. The 1991 Honeydew earthquake occurred at depths of 8 to 10 km (Figure 4). Most incoming sediments on the Gorda plate are accreted to the northern California margin [20]. [21] measured P and S velocities near the hypocenter of the 1992 shock. They also indicate the 1991 and 1992 thrust events occurred within the North American plate and not on the plate boundary. The region is unusual in that some of the youngest oceanic crust world-wide is entering the subduction zone west of the 1992 rupture zone.

Where slip occurred in 1992 has possible implications for great earthquakes to the north along the Cascadia subduction zone. [18] indicated that a portion of the old Farallon plate was accreted to the North American plate in the region of abundant seismicity between 40.3˚N and 41.2˚N in Figure 1. If so, the implications for great shocks farther north are questionable except for their being thrust mechanisms on shallow-dipping faults.

The northwestern extent of the San Andreas fault is shown in Figure 1. The change in its strike shown near 40˚N, 124.1˚W [18], if correct, indicates greater structural complexity in the area of the 1992 shock. How far the 1906 great earthquake may have ruptured into that area and the triple junction is uncertain. [22] modelled geodetic and seismic data near the northwestern end of the 1906 rupture zone. They conclude that large slip in 1906 occurred as far northwest as Shelter Cove (Point Delgada) near 40˚N just to the south of the Mendocino region. The San Andreas fault comes ashore at Point Delgada before going offshore again to the northwest [23] [24].

[25] concludes that large earthquakes near the southern end of the Cascadia subduction zone may be linked with great events along the northernmost San Andreas fault. Their dates are uncertain, however, since they are from paleoseismic measurements. [25] indicates that southern Cascadia has a short repeat time of about 220 years, shorter than those of subduction segments to the north.

The Uniform California Earthquake Rupture Forecast [26] assigned the offshore Noyo segment of the northern San Andreas fault near 39.5˚N in Figure 1 a mean repeat-time interval of 188 years (with one standard deviation of about 25 years). Since it last ruptured in a great earthquake in 1906, it is unlikely to rupture again before the year 2069. The Mendocino segment that broke in 1992 seems unlikely to rupture before then even if it broke in 1906.

Figure 8 shows the distances of forerunning earthquakes and their times before the 1992 Cape Mendocino and 1989 Loma Prieta mainshocks. The short-period hypocenters, i.e., their rupture initiations, were used in determining those distances. The Lake Elsman precursors to the 1989 mainshock occurred on a separate fault zone that was very close to the 1989 rupture zone only 15 months earlier [27]. The 1991 Honeydew shock also took place very near the 1992 hypocenter 8 months beforehand.

Stress dropped in the great 1906 California mainshock along the segment of the San Andreas fault that broke in 1989. A similar drop in stress and its subsequent buildup by plate motion are taken to have occurred for the 1992 mainshock. The occurrence of shocks along the Calaveras fault zone about 30 km from the 1989 rupture zone starting a decade earlier may have been a long-term

(decadal) precursor [27] [28]. The occurrences of those nearby precursors to the 1989 and 1992 shocks are taken to be especially important for intermediate (yearly)-term forecasting. I urge scientists to be alert for other events of significant magnitude close to active faults since they may be intermediate-term precursors of a soon-to-occur larger earthquake.